Inhibition of transforming growth factor-|[beta]|-activated kinase-1 blocks cancer cell adhesion, invasion, and metastasis

Tumour cell metastasis is a complex, multi-step process involving cell adhesion, invasion, and migration on extracellular matrix (ECM). During this process, cell signalling is critical for the interaction of tumour cells with other cells, the ECM, and components within tumour microenvironments (Talmadge and Fidler, 2010). We have focussed on defining and characterising pathways (Paine et al, 2000; Palmantier et al, 2001) that are activated by polyunsaturated fatty acids found both in typical Western diets (Rose and Connolly, 1997) and in response to inflammatory signals (Honn et al, 1992). We hypothesise that one critical pathway involves TGF-β-activated kinase-1 (TAK1) (Nony et al, 2005). TGF-β-activated kinase-1 is a member of the mitogen-activated protein kinase (MAPK) family, and has been implicated in various signalling pathways (Yamaguchi et al, 1995; Ninomiya-Tsuji et al, 1999). The downstream targets of TAK1 include MKK and IKK, which in turn can activate p38 MAPK, JNK, or NF-κB signalling pathways (Landstrom, 2010). Thus, TAK1 regulates a number of biological processes including immune responses, stress responses, and inflammation (Landstrom, 2010). Tumour cells can be strongly affected by components of the extracellular microenvironment (Talmadge and Fidler, 2010), including the n-6 polyunsaturated fatty acids (PUFAs) such as arachidonic acid (AA) and its precursor linoleic acid. Both fatty acids are consumed in substantial amounts in Western diets. In tumour cells, PUFAs have been shown to stimulate an adhesive and invasive phenotype in breast cancer cells in vitro and to promote tumour growth and metastasis in animal models (Rose et al, 1991; Connolly and Rose, 1993; Palmantier et al, 1996). Fatty acids induce various signal transduction pathways in MDA-MB-435 cells, a highly metastatic human breast cancer cell line frequently used in models of breast cancer (Hurst et al, 2009). This cell line is particularly useful for the study of metastasis, given its preference for growth in a mammary environment (Bao et al, 1994), and displays an aggressive, metastatic phenotype accompanied by the expression of both mammary marker proteins and melanocyte-related genes, similar to patterns observed in freshly resected, human breast tumours (Montel et al, 2009). Activation of these cells with fatty acids in vitro leads to β1 integrin activation and increased cell adhesion to collagen IV (Palmantier et al, 2001). Arachidonic acid activates a p38 MAPK pathway that is required for adhesion to collagen IV (Paine et al, 2000; Garcia et al, 2009). Thus, in vitro, these tumour cells take on many of the characteristics of highly metastatic cells, consistent with the increased metastasis seen in animal models in which mice are fed high levels of dietary n-6 PUFAs (Rose and Connolly, 1997). Preliminary evidence from our laboratory suggested that TAK1 may have a role in AA-induced p38 activation (Nony et al, 2005). Herein, we investigated the role of TAK1 in cancer cell adhesion and invasion in vitro, and in tumour growth and metastasis in vivo. Invasion of cells through Matrigel- (BD Biosciences, Franklin Lakes, NJ, USA) (1 : 25 dilution) coated nucleopore membranes (8 μm pores) (Fisher Scientific, Pittsburgh, PA, USA) was performed in Boyden chambers (Neuro Probe, Gaithersburg, MD, USA) for 1 h at 37 °C. Ten percent FBS was placed in the bottom chamber and 2.5 × 104 cells were placed in the top chamber with 1% FBS medium for 24 h. The invaded cells were stained with Diff-Quik (Dade Behring Holdings Inc., Deerfield, IL, USA) and the cells were counted in four fields per membrane on triplicate membranes. Next, the parental MDA-MB-435, control shRNA, and the two TAK1 shRNA cell lines were tested for adhesion to collagen IV and invasion through Matrigel. Control shRNA and parental cells had similar adhesion responses, with AA inducing a three-fold induction of cell adhesion as compared with the control (Figure 3). In contrast, the TAK1 knockdown cells were unable to increase their adhesion to collagen IV in response to AA treatment, demonstrating the requirement for TAK1. Control shRNA cells and parental MDA-MB-435 cells demonstrated a significant increase in invasion when treated with AA when compared with the vehicle-treated control (Figure 4A). However, shRNA knock-down of TAK1 completely blocked the AA-induced invasion in both TAK1 shRNA cell lines (Figure 4B). We used tumour size, instead of time following injection, as the criterion for resection, because tumour size is strongly dependent on angiogenesis, which is required for growth and which provides a route for tumour cells to escape the original site (i.e., haematogenous metastasis) (Folkman, 2002; Weinberg, 2007). As most tumours spread by haematogenous metastasis, allowing the tumours to reach a similar size should allow for a more equivalent opportunity for metastasis. The TAK1 shRNA tumours were removed at week 8 and the control shRNA tumours were removed at week 9 following injection, at which time the tumours had equivalent volumes (Figure 5C). After primary tumour resection, the mice were imaged weekly for metastatic tumours. In the control group, lung metastases were detectable as early as 1-week post-resection (Figure 6A). There were distinct luminescent foci that appeared to be lung tumours. The per cent of control mice with metastasis increased steadily over the 7 weeks (Figure 6B). As depicted in Figure 6A, a few of the TAK1 shRNA mice developed metastases, but many of the mice were negative (e.g., mouse #4). In fact, as seen in Figure 6B, only 38% of mice in the TAK1 shRNA group had metastases compared with 83% for the control shRNA group at 7 weeks post resection. Additionally, the number of control shRNA mice that developed metastases steadily increased over time. However, the number of TAK1 shRNA mice with metastases only increased slightly over the 7-week time course. Overall, the number of mice that developed metastases over time was significantly higher in the control group compared with the TAK1 shRNA group (P=0.004 by log-rank test). The trend over time revealed that the TAK1 shRNA group had a significantly lower luminescence (Supplemental Figure 4A). Futhermore, more animals in the control group displayed evidence of metastasis as early as 2 weeks after the tumour resection. Thus, it is unlikely that the 1-week difference in time that the animals carried the primary tumour resulted in the striking difference in metastasis observed between the two groups. This study demonstrates that TAK1 is a critical component for the induction of increased adhesion, invasion and metastasis. Most importantly, this is the first report demonstrating that TAK1 knockdown in an orthotopic model of spontaneous breast cancer metastasis results in a significant reduction in metastasis frequency and a striking decrease in lung tumour burden. In the present study, we found that the MDA-MB-435 control shRNA cells formed tumours in the mammary fat pad and were highly metastatic to the lung in animals on the normal NIH-31 rodent diet. These numbers are comparable to what is described for the MDA-MB-435 cells (Price et al, 1990). We did not detect a higher rate of metastasis in either the control or TAK1 shRNA tumour-bearing animals when they were maintained on a 12% linoleic acid diet (data not shown). This result was not surprising because we observed such a high metastatic rate in the control animals on the diet containing 4.6% linoleic acid. Therefore, the findings from our metastasis experiments are more broadly applicable in that the role of TAK1 is not likely limited to metastasis induced by high levels of fatty acids but is important in pro-metastatic signalling pathways initiated under relatively normal dietary conditions. Interestingly, the TAK1 shRNA cells established as a tumour in vivo slightly faster than the control shRNA cells, even though, in vitro, there was no difference in growth between these two cell lines. However, we observed that the control shRNA tumours, although slower to establish initial growth, rapidly increased in size once the tumours were of palpable size. Even though, this initial difference in tumour growth was observed, the TAK1 shRNA tumours were significantly less likely to metastasise than the control shRNA tumours. Any number of factors could have influenced the initial growth of the tumour cells in vivo that would not have been present in vitro. It is possible that other pathways were affected by TAK1 knockdown that could have resulted in a different response of these cells when exposed to the extracellular environment of the mammary fat pad that allowed them to establish into a tumour faster than the control shRNA cells. This finding is in contrast to that reported in a different cell line in which a dominant negative TAK1 construct inhibited tumour growth in vivo (Safina et al, 2008). Thus, although inhibition of tumour growth may also block metastasis, our studies reveal that TAK1 is having an important role in the successful metastasis of breast tumour cells even in the presence of normal, or slightly enhanced, growth in the mammary fat pads, suggesting that inhibition of TAK1 through small molecules or dietary changes may be able to block metastasis specifically. TGF-β-activated kinase-1 is an important signalling molecule that activates numerous signal transduction pathways, many of which are associated with cancer. For example, TAK1 is a key upstream kinase in the NF-κB pathway, which promotes cell survival and proliferation and which has been the focus of treatment intervention for many types of cancers. We previously showed that lysine-63-linked ubiquitination, a pathway that involves activation of signalling but does not lead to proteasomal degradation, was required for AA to induce cell adhesion (Ray et al, 2010). In contrast, the lysine-48-linked ubiquitin-mediated degradation pathway was not required for AA-mediated cell adhesion (Ray et al, 2010), suggesting that NF-κB, which requires lysine-48 ubiquitination, does not participate in fatty acid-induced signal transduction. We do not currently understand the mechanism by which AA exposure leads to TAK1 activation, although we have examined a number of possibilities. For example, a combination of dominant negative and siRNA experiments have shown that at least TRAF 2, 5, and 6 are not responsible for activation of TAK1 (data not shown). Furthermore, both TAB1 and TAB2 are associated with TAK1 in these cells both in the presence and absence of AA (data not shown). We have previously shown that inhibition of metabolism of AA to 15S-HETE completely blocks cell adhesion and p38 phosphorylation (Palmantier et al, 1996; Nony et al, 2005); thus, interactions between this metabolite and specific kinases or their binding partners would provide interesting targets for inhibiting these processes and is consistent with previous studies of the importance of HETEs in tumour cell behaviour (Szekeres et al, 2002). We are currently investigating the biochemical interactions of 15S-HETE with various signalling proteins to determine whether this molecule is directly responsible for kinase activation. Our results demonstrate that TAK1 expression in these tumour cells is critical for key steps in tumour cell metastasis; for example, TAK1 knockdown is associated with a decrease in the ability of the tumour cells to invade through and attach to key ECM proteins. These results provide a link between the effects of the loss of TAK1 in the tumour cells in vitro, and a clear demonstration of both fewer animals developing metastases and reduced metastatic burden in the lung. On the basis of the in vitro assays, these tumour cells lose their ability to adhere to and invade the ECM in response to specific external stimuli, in this case, AA that are commonly found in the tumour microenvironment. Thus, these tumour cells apparently require TAK1 to effectively escape the primary tumour and successfully colonise a secondary site. All together, these findings demonstrate that TAK1 is an important pro-metastatic signal transducer and represents a potential target for anti-metastasis therapy. The authors declare no conflict of interest. We thank Dr Ron Herbert for the pathology analysis, Dr Shyamal Peddada for statistical analysis, Norris Flagler for image analysis, and the Histology Core Facility at NIEHS for help with necropsy and tissue preparation. We thank Drs Darlene Dixon and Robert Langenbach for a careful reading of the manuscript. This work has been supported in part, or in whole, by the Intramural Research Program of the NIEHS and the NIH. Supplementary Information accompanies the paper on British Journal of Cancer website